[m-dev.] cvs diff: Bug fix for layoutst

[Fergus Henderson]

Tyson Richard DOWD, you wrote:
> 
> > I have been using the name `type_of' for the following function:
> > 
> > 	:- type (type).
> > 		% `type' is another name for `type_info(T)'.
> > 		% ... some low-level magic to implement unification (etc.)
> > 		% of values of type `type' omitted ...
> > 
> > 		% type_of(X) = T iff T is the type of X.
> > 	:- func type_of(T::unused) = ((type)::out) is det.
> > 	:- pragma c_code(type_of(_X::unused) = (Type::out),
> > 			will_not_call_mercury, "Type = TypeInfo_for_T;").
> 
> Then you've changed your names slightly from when you asked for this to
> be implemented - you used type_info, not type. Otherwise it is much
> the same.

OK, good.  The name type_info/0 is fine.

> The c_code basically does what you ask, but will also need to check for
> `univ' and give an error or something, (you can't do anything with a
> univ without some data),

Uh, why not?

If we had existential types, we could implement univ as follows:

	:- module my_univ.
	:- interface.
	:- type univ.
	:- pred type_to_univ(T::in, univ::out) is det.
	:- pred univ_to_type(univ::in, T::out) is semidet.

	:- implementation.

	:- type univ ---> some [T] univ(T).
	type_to_univ(X, univ(X)).

	univ_to_type(univ(X), Y) :-
		(if type_of(X) = type_of(Y) then
			Y = unsafe_cast(X)
		else
			fail
		).

Although we currently don't have existential types, it would make sense for
`univ' to behave as if it were implemented in this way.

-- 
Fergus Henderson <fjh at cs.mu.oz.au>   |  "I have always known that the pursuit
WWW: <http://www.cs.mu.oz.au/~fjh>   |  of excellence is a lethal habit"
PGP: finger fjh at 128.250.37.3         |     -- the last words of T. S. Garp.